Claude-Code
Claude Code Hands-On (10): Skills, and When to Reach for Each Extension Mechanism
Skills are the newest extension mechanism in Claude Code: a folder, a SKILL.md, and a body of instructions the model loads on demand. How they differ from slash commands, MCP servers, and hooks — with a decision tree for …
Claude Code Hands-On (9): settings.json, the Three-Layer Permission Model, and Env
settings.json is the file that decides what Claude can do, where, and with whose credentials. The three layers (user, project, local), the permission grammar, env vars that change behavior, and the precedence order that …
Claude Code Hands-On (8): Sub-Agents, Worktrees, and Plan Mode
Three features that change what Claude Code can take on at once: sub-agents for parallel research, worktrees for isolation, plan mode for the moments before you let it touch anything. The boundaries between them, and …
Claude Code Hands-On (7): Ten Hooks I Actually Use, with the Code
Picking ten hooks out of the 100 in the reference repo and walking through each: what it does, the actual JS, the settings.json wire-up, and where it bites. PreToolUse for safety, PostToolUse for hygiene, the boring ones …
Claude Code Hands-On (6): The SDK, GitHub Integration, and Claude in CI
The SDK turns Claude Code from a CLI into a library. GitHub Action makes it answer @claude on PRs. Together they let you put Claude inside the same CI pipeline that already runs your tests — without giving up control.
Claude Code Hands-On (5): Hooks, or How to Stop Worrying About Yolo Mode
Hooks are the shell scripts that run before and after every tool call. PreToolUse can block. PostToolUse can format, lint, log. Five hooks I use on every repo, and the one anti-pattern that bites everyone.
Claude Code Hands-On (4): MCP Servers, or How Claude Talks to Anything
MCP is the plug-in protocol that lets Claude Code reach beyond your filesystem. Install one (Playwright), see it work end-to-end, and learn the permission model that keeps it from going feral.
Claude Code Hands-On (3): Custom Slash Commands and Conversation Control
Slash commands turn repeated workflows into one-line invocations. $ARGUMENTS makes them parameterized. The right ones become your team's shared vocabulary.
Claude Code Hands-On (2): Shortcuts, the Four-State Toggle, and Thinking Modes
Shift+Tab is a four-state cycle, not a binary. Thinking modes have five levels. Escape and double-Escape do different things. Five shortcuts that reshape how Claude Code feels day-to-day.
Claude Code Hands-On (1): Install, the Three-Layer Config, and the # @ /init Trio
Install Claude Code, understand the three-layer settings.json system, and learn the three quietly powerful primitives: # to write into context, @ to reference files, /init to generate the project memory file.